Accessing several HP EVAs from the same ESX 3 server

[paulo_meireles]

Hi,

We have an HP EVA 5000 that is showing its age. It had firmware 3.028, and it was s-l-o-w. After all VMs were migrated to an EVA 6000 (through the network) the EVA 5000 was upgraded to firmware 4.004. It's much faster now - most of its load was moved to the EVA 6000... :smileysilly:

I'm now being asked if we may have some VMs on the EVA 5000 and some on the EVA 6000. I seem to recall something about ESX not supporting visibility to more than one EVA at a time, but I cannot find it anymore.

So, my questions are:

\- may we have several ESX servers seeing VMFS volumes on both the EVA 5000 and the EVA 6000 at the same time?

\- if we had another EVA 6000, could we have the same ESX servers seeing both?

Regards,

Paulo

[Jae_Ellers]

We're running luns on two EVA 8000. All our enterprise ESX hosts are connected to some of these luns on both EVAs with no issues.

[VirtualKenneth · Accepted solution]

Same here, I'm serving LUN's from my EVA5 and EVA4 to ESX simultaneously and preparing a site which will run active/active over two EVA4's on 2 sites with CA between them.

[bggb29]

you need to make sure you do not try to use the same lun id's.

Start 1 eva at 1 and the other at 100 or 200 for lun ids

[Jae_Ellers]

you need to make sure you do not try to use the same

lun id's.

Start 1 eva at 1 and the other at 100 or 200 for lun

ids

What's wrong with using the same lun ids? We do this and have no issues.

The SP ID and the lun ID are used to create the lun signature.

Please point me to the source of this statement.

[bggb29]

we had problems, they may not be an issue anymore, We where either on 1.5 or 2.x at the time.

it was advised by our san installation consultants, they had us setup using the second san with a increment of 100.

[BUGCHK]

What's wrong with using the same lun ids? We do this and have no issues.

It works, because the second storage array makes up a complete new set of SCSI targets.

However, I would make them unique if there is a slight chance the Continuous Access (CA) is introducted now or in the future.

If you have duplicate LUN addresses, you would have to present the destination virtual disk on a different address. This definitely requires a resignature to make the VMFS available again, which can mean quite some work: you have to re-register all VMs on that VMFS because the path down from /vmfs/volumes/ changes.

I had to do it last week.

[Jae_Ellers]

Good point. We might use CA in the future.

I still don't think this is an issue because the signature is based on the storage processor and lun ID. You should never be in a situation where your live lun and mirror lun are presented to the same host at the same time without requiring resignaturing.

[BUGCHK]

I agree.

CA even allows you to set up the virtual disk presentation on the source AND the destination vdisk. By default, the destination is not visible by the host unless a failover is done or some special setting changed.

That's quite handy in case you don't have any pre-tested failover scripts.

[paulo_meireles]

Thanks for your answers. So, it seems not to be a problem to present LUNs from two different EVAs to the same ESX host. Somehow, it itches me to be unable to locate the information regarding the impossibility to do so. Could it be that we should not have an active/passive EVA and an active/active EVA connected to the same ESX host? I'm starting to doubt my memory...

Paulo

[Jae_Ellers]

You can present them, but if one lun is a copy of another lun and you have your settings configured to allow snapshot luns and disable resignature then you will have one confused ESX server and corruption may occur.